Writing Fights, Crowds and Group Scenes for AI Video

Quick answer: AI video does best when each shot has a few named people and one clear action. Write a fight as a run of short shots, one blow or one reaction each, rather than one shot of a whole brawl. Put a crowd in the segment’s background-people line and keep it behind the named characters. For a group scene, show two or three people per shot and cut between them instead of framing everyone at once. Show the moment before or after a violent act rather than the act in detail, which also keeps clear of content review.

Still from “Rust City Knockout”, an AI short drama generated end-to-end by SceneMixer (6 characters, 1 episode) — Sample: “Rust City Knockout” — Painterly Anime, 1 episode, produced end-to-end by SceneMixer from the script · watch the episodes

Find the busiest shot in your episode, count the named people in it, and split it if there are more than three.

Fights: one blow per shot

Break the fight into beats: the swing, the block, the face of the one who takes the hit, the fall. Give each its own shot of two or three seconds and name who does what with @. Name the weapon or object with @ in the shot where it is used, so it keeps its look. Wide shots of many people fighting at once tend to blur; the cut between close shots carries the energy. How to write one shot · Props.

A storyboard segment: location, time and weather fields, then numbered shots with durations, asset chips and spoken lines — A storyboard segment: each numbered shot carries one action and its own seconds.

Injuries and violence

Show the cause and the result, not the process: the raised knife, then her hand pressed to her side. Detailed wounds and gore are likely to be refused by content review. Blocked by moderation.

Crowds

A crowd is not a list of characters. Write it once in the segment’s background-people line — who, where, doing what — and keep the named characters in front. Anyone who speaks or is looked at individually is a character, not part of the crowd. Background people.

Group scenes: the dinner of six

Four or more named characters in one shot is where faces start to blend. Open on one wide shot if you need the room, then cut to pairs: the host and the guest who answers, the two who exchange a look. Name only the people in frame in each shot; the others are off screen for that moment. My character’s face changed · Who is in frame.

Big set pieces

Battles, stadiums, car chases with many vehicles: suggest the scale and film the people. A close shot of a soldier with smoke and noise behind him says “battle” better than an army. Sound helps: write it in the shot’s Ambient line.

Scene	Weaker	Stronger
Fight	One 10-second shot of a brawl	Four short shots: swing, block, hit, fall
Crowd	A list of twenty extras as characters	One background-people line, named leads in front
Dinner of six	Everyone in every shot	A wide opener, then pairs
Battle	An army in a wide shot	One soldier, smoke and sound

FAQ

How many people can be in one shot?

Two or three named people is the comfortable range. With four or more, faces start to blend.

Can I have a long fight in one take?

It is possible, but long takes of fast action tend to drift. Short shots with clear actions hold up better.

Why did my fight get blocked?

Detailed injury or gore. Show before and after instead.

Do extras need sheets?

No. Background people come from one line in the segment header.

Few people, one action, many cuts

Split the busy shots and let the cuts carry the scene.

Open the editor

Updated October 2, 2026